Reducing memory bandwidth is of utmost importance, such as when designing a graphics architecture, as energy efficiency is a performance factor that weighs most heavily in hardware design. For example, transactions over memory buses may cost several orders of magnitude more than computation in terms of energy and latency. Hence, it is common to attempt to reduce bandwidth usage at the expense of more computations and thus reducing power consumption and/or increasing overall performance. This serves as a motivation behind buffer compression algorithms, commonly found in a graphics processing unit (“GPU” or “graphics processor”).
For example, ray tracing is becoming increasingly important in today's rasterization-based central processing unit (“CPU” or “application processor”) for complementing rasterization in achieving various graphics processing-related tasks, such as generating pixel-exact shadows. Hence, compressions of all sorts are regarded as important in continuously achieving better performance in graphics processors.